Canon 17mm f/4L Tilt-Shift Lens Review

[Brian Hirschfeld]

Wrote a quick article a little while back on some use of mine of the Canon 17mm f/4L Tilt-Shift lens and never published it. Just published it with an update about the announcement of the 5Ds and 5Ds R.

I think that its an interesting part of the ultra high-resolution 35mm DSLR debate because of the fact that it does start to entire "medium format" territory and its interesting to consider the place of these cameras as well as the lenses that are available for them. We are all aware that the Canon 17mm can be used up to 80mp as myself and many others have shown.

http://brianhirschfeldphotography.com/2015/02/11/canon-ts-e-17mm-f4l-tilt-shift-lens-brief-review/

[Gary Ferguson]

Brian, thanks for posting. I hope you don't take offence but I thought the illustrative shot you used was "over compensated", which to my eye at least is more intrusive than the converging verticals we'd expect from a non-shifting 17mm lens. Some of the architectural photographers I most respect often leave just a little element of converging verticals, even when shooting with technical cameras, as they feel full correction is too artificial and brutal a look, and by implication over compensation would be even more so.

[Slobodan Blagojevic]

Gary, if I understood you correctly, "over compensated" means that when verticals are made parallel, that is already too much compensation (and I agree with that). In Brian's case, the compensation is actually beyond that point, as it goes in the opposite direction, quite visible on the left side of the image. Not that any of that detracts from the rest of his article.

[Phil Indeblanc]

Nice, I've been wanting to get either of the 3 flavors and less so for arctitecture, but the 90 is pretty old and was hoping for an update to the lens...But I'm glad to read it.

The image may have some lines going in at the bottom, but I think the palm trees are what "throw off" the eyes... more than the buildings verticals.

(keep in mind, good MFDb like the PhaseOne started at 22mpixels)

[Phil Indeblanc]

I agree. rulers do not replace eyes. 

[pfigen]

It might be true that rulers don't replace eyes, but anyone looking at Brian's posted image who can't see with their own two eyes how far off that image is, well, they either need a new pair of eyes or a ruler.

The right side of the building is straight and everything else tips off to the left. That particular building is not the type of image where you might purposefully under correct. Just a mediocre example that says very little about the lens in review.

[Phil Indeblanc]

So far I think we all see it.
Perhaps Paul is simply mentioning something anecdotal that we can use in general.
I don't think anyone is saying the image looks correct.

I certainly value my measuring devices, and this image does look like it tapers in below left. I think the palm tree shape exaggerates this visual.
(No rulers used in observation)


Here it is open in PS to demo....

[pfigen]

Your screen shot shows the tilt. It's a tricky image because the architectural details on the right side also give the impression, first off, of a good dose of pincushion distortion, something I've never encountered with this lens. This is a great example of why guides in Ps are so important. It's almost always best to start off with the image straight according to the guides and then deviate from there, but this one was never straightened. Worse, probably the camera was never leveled.

[Bart_van_der_Wolf]

We have all read numerous times that the 17 & 24 TS-E lenses ( or any ultra wide) are brilliant lenses for architecture. I don’t think they are, most of the time I see them as necessary evils. They let  the scene be captured but there is no need to accept their exaggerated view when it looks wrong because we have Ps to regain control of the image.

Hi Paul,

That's not the fault of the lenses. It is caused, assuming correct keystoning correction, by projection of the image on a flat plane and viewing that result from too far away and from the wrong position (not from the virtual center of projection).

The projection near the edge of the image circle on a flat plane will seemingly stretch the image more the closer it gets to the edge, and when we view the output from the wrong position (too far away) we will 'enhance' the stretched sensation. However, if we were to view the output from below and from very (uncomfortably) close distance, everything would look perfectly normal.

So, to cope with our 'wrong' viewing conditions, we can a.) under-correct the keystone correction, and/or b.) compress the image in the shift direction, and/or c.) use a larger output size.

A Raw converter like Capture One, typically under-corrects the keystone correction by a user definable percentage, and offers the option to compress it by changing the aspect ratio. Pano stitching software, which may offer a higher quality result than a wider angle lens, allows to reposition the center of projection (even on a single image), and use an offset and or aspect ratio changes to influence the apparent 'look' of the resulting image.

Cheers,
Bart

[BobDavid]

Palm Beach? Anyway, my eyes go loopy when I see the picture. Or is the building loopy? Are we all loopy?